: tax salience and tax rates

E-ZTAX: TAX SALIENCE AND TAX RATES∗

AMY FINKELSTEIN

This paper examines whether the salience of a tax system affects equilib-rium tax rates. I analyze how tolls change after toll facilities adopt electronic tollcollection (ETC); drivers are substantially less aware of tolls paid electronically. Iestimate that, in steady state, tolls are 20 to 40 percent higher than they wouldhave been without ETC. Consistent with a salience-based explanation for this tollincrease, I find that under ETC, driving becomes less elastic with respect to thetoll and toll setting becomes less sensitive to the electoral calendar. Alternativeexplanations appear unlikely to be able to explain the findings.

I. INTRODUCTION

For every dollar of revenue raised by the U.S. income taxsystem, taxpayers incur about ten cents in private compliancecosts associated with record keeping and tax filing (Slemrod 1996).These compliance costs impose a deadweight burden on society.Yet policies that would reduce these costs are frequently opposedby policy makers and economists who believe that compliancecosts play an important role in keeping taxes visible and salient tothe electorate, who then serve as an important check on attemptsto raise the scale of government activity beyond what an informedcitizenry would want.

For example, Milton Friedman has publicly lamented his in-advertent contribution to the growth of government by encourag-ing the introduction of the visibility-reducing Federal income taxwithholding system during the Second World War (Friedman andFriedman 1998, p. 123). More recently, in 2005, the President’sAdvisory Panel on Federal Tax Reform failed to reach consensuson whether to replace part of the existing income tax system witha value-added tax (VAT), in part because of concerns about how

∗I am grateful to Daron Acemoglu, Gene Amromin, Pol Antras, David Autor,Raj Chetty, Peter Diamond, Liran Einav, Hanming Fang, Naomi Feldman, Ed-ward Glaeser, Mike Golosov, Austan Goolsbee, Jerry Hausman, Larry Katz, ErzoLuttmer, Brigitte Madrian, Sean Nicholson, Ben Olken, Jim Poterba, Nancy Rose,Stephen Ryan, Monica Singhal, Heidi Williams, Clifford Winston, two anonymousreferees, and seminar participants at Cornell, MIT, Berkeley, Stanford GSB, Yale,the NBER Public Economics Meeting, Harvard, and Stanford for helpful com-ments; to James Wang and especially Julia Galef for outstanding research assis-tance; to Tatyana Deryugina, Julia Galef, Stephanie Hurder, and Erin Strumpf forhelp in conducting the survey of toll awareness; and to the innumerable employeesof toll operating authorities around the country who generously took the time toprovide data and to answer my many questions.

C© 2009 by the President and Fellows of Harvard College and the Massachusetts Institute ofTechnology.The Quarterly Journal of Economics, August 2009

969Downloaded from https://academic.oup.com/qje/article-abstract/124/3/969/1905173by Mass Inst Tech useron 13 August 2018

970 QUARTERLY JOURNAL OF ECONOMICS

the lower visibility of a VAT would affect the size of government.As the Advisory Panel noted in its report:

[Some] Panel Members were unwilling to support the [VAT] proposal giventhe lack of conclusive empirical evidence on the impact of a VAT on the growthof government. Others were more confident that voters could be relied on tounderstand the amount of tax being paid through a VAT, in part because theproposal studied by the Panel would require the VAT to be separately statedon each sales receipt provided to consumers. These members of the Panelenvisioned that voters would appropriately control growth in the size of thefederal government through the electoral process. (The President’s AdvisoryPanel on Federal Tax Reform 2005, pp. 203–204)

The idea that a less visible tax system may fuel the growthof government can be traced back at least to John Stuart Mill’s1848 Principles of Political Economy. It has its modern roots inthe public choice tradition of “fiscal illusion.” In a series of influ-ential books and articles, James Buchanan and co-authors haveargued that citizens systematically underestimate the tax price ofpublic sector activities, and that government in turn exploits thismisperception to reach a size that is larger than an informed citi-zenry would want. The extent of the tax misperception—and thusthe size of government—is in turn affected by the choice of taxinstruments, with more complicated and less visible taxes exacer-bating the extent of fiscal illusion and thereby increasing the sizeof the government (e.g., Buchanan [1967]; Buchanan and Wagner[1977]; Brennan and Buchanan [1980]).

Empirical evidence of the impact of tax salience on tax rates,however, has proved extremely elusive. Most of the evidence comesfrom cross-sectional studies of the relationship between the sizeof government and the visibility of the tax system, where thedirection of causality is far from clear (Oates 1988; Dollery andWorthington 1996). Moreover, as I discuss in more detail below,the sign of any effect of tax salience on tax rates is theoreticallyambiguous. The link between tax salience and tax rates is there-fore an open empirical question.

In this paper, I examine the relationship between tax salienceand tax rates empirically by studying the impact of the adoption ofelectronic toll collection (ETC) on toll rates. Electronic toll collec-tion systems—such as the eponymous E-ZPass in the northeasternUnited States, I-Pass in Illinois, or Fast-Trak in California—allowautomatic deduction of the toll as the car drives through a tollplaza. Because the driver need no longer actively count out andhand over cash for the toll, the toll rate may well be less salient

Downloaded from https://academic.oup.com/qje/article-abstract/124/3/969/1905173by Mass Inst Tech useron 13 August 2018

E-ZTAX: TAX SALIENCE AND TAX RATES 971

to the driver when paying electronically than when paying cash.Indeed, I present survey evidence that indicates a strikingly lowerawareness of the amount paid in tolls by those who pay electron-ically relative to those who pay using cash. This discrepancy intoll awareness exists even among regular commuters on a tollfacility. As a result, toll facilities’ adoption of ETC—and the re-sultant switch by many drivers to paying electronically—providesa setting in which to examine the impact of tax salience on taxrates.

Different toll facilities in the United States have adoptedETC at different points in time over the last several decades,and some have not yet adopted it. To study the impact of ETC,I examine the within toll-facility changes in toll rates associatedwith the adoption and diffusion of ETC. To do so, I collected a newdata set on the history of toll rates and ETC installation for 123toll facilities in the United States. Where they were available, Ialso collected annual facility-level data on toll traffic, toll revenue,and the share of each that is paid by electronic toll collection.

I find robust evidence that toll rates increase after the adop-tion of electronic toll collection. My estimates suggest that whenthe proportion of tolls paid using ETC has diffused to its steadystate level of about 60 percent, toll rates are 20 to 40 percenthigher than they would have been under a fully manual toll col-lection system.

I also present evidence of two potential mechanisms by whichreduced salience may contribute to increased toll rates. First, Ifind that the elasticity of driving with respect to the toll declines(in absolute value) with the adoption of electronic toll collection,suggesting that ETC may raise the optimal level of the toll. Sec-ond, I show that under ETC, toll-setting behavior becomes lesssensitive to the local election calendar, suggesting that ETC mayreduce the political costs of raising tolls.

The rest of the paper proceeds as follows. Section II providesa conceptual framework for how tax salience may affect tax ratesand the factors that may affect the (ambiguous) sign of this rela-tionship. Section III presents evidence that tolls are less salientwhen paid by ETC than by cash. Section IV describes the dataon toll rates and driving. Section V estimates the impact of ETCon the elasticity of driving with respect to the toll. Section VIestimates the impact of ETC on toll rates. Section VII considersnon-salience-based explanations for these empirical findings. Thelast section concludes.



II. EFFECTS OF TAX SALIENCE ON CONSUMERS AND GOVERNMENT:CONCEPTUAL FRAMEWORK

In a fully salient tax system, individuals are aware of actualtaxes as they make economic and political decisions. In a lesssalient tax system, individuals are not aware of the actual tax(τ ), but instead have a perception of the tax, which I denote byτ . Recent empirical evidence is consistent with individuals mis-perceiving taxes (Liebman and Zeckhauser 2004; Feldman andKatascak 2005; Chetty, Kroft, and Looney forthcoming) and withthe salience of the tax affecting the extent of this misperception(Chetty, Kroft, and Looney forthcoming).

This paper focuses on the response of tax rates to tax salience.However, because an input into this response is how consumers’economic behavior is affected by tax salience, I begin—in both theconceptual framework and the subsequent empirical work—by an-alyzing the consumers’ response; I then turn to the government’sresponse.

I denote by θ ≥ 0 the (lack of) salience of the tax system. Ahigher θ corresponds to a less salient tax system; θ = 0 corre-sponds to a fully salient system. In the empirical application Iwill examine the move from manual (i.e., cash) toll collection toelectronic toll collection (ETC) and interpret this as a move toa less salient tax system (i.e., an increase in θ ); I present surveyevidence in Section III that is consistent with the assumption thatETC reduces the salience of tolls.

There are two types of tax salience that may affect tax setting:tax salience at the time of the consumption decision for the taxedgood, and tax salience at the time of voting. These need not be thesame. To capture this, I denote the perceived tax by τ j , where j ={c, v} indicates perceived taxes at the time of consumption and ofvoting, respectively.

For simplicity I assume the perceived tax is a linear functionof the actual tax,

(1) τ j(θ ) ≡ δ0 j(θ ) + δ1 j(θ )τ,

and normalize a fully salient system as one in which the per-ceived and actual tax are the same (i.e., δ0 j(0) = 0 and δ1 j(0) = 1).I assume that δ1 j(θ ) > 0 (i.e., the perceived tax is increasing inthe actual tax). I also assume that in a less salient tax system,the link between the perceived and the actual tax is weaker (i.e.,δ′

1 j(θ ) < 0). The effect of the tax salience on the perceived toll level



is, however, a priori ambiguous; in other words, δ′0 j(θ ) can be

either sign. For simplicity, I consider only cases of positive tax-ation (τ > 0), and further assume that τ j > 0.

II.A. Response of Consumer Economic Behavior to Tax Salience

The individual chooses consumption of the taxed good basedon the perceived tax at the time of the consumption decision, τC(θ ).To simplify the analysis, I assume the individual maximizes autility function that is quasi-linear in the taxed good and exhibitsconstant elasticity of demand.1 The individual thus solves

(2) maxx1

γ0x( 1

γ1+1)

1 + x2 subject to x2 + (p + τC(θ ))x1 ≤ m,

where x1 denotes the taxed good (with producer price p), x2 denotesall other goods (whose price has been normalized to 1), and m isconsumer income. I denote by η(τC) ≡ γ1 the (constant) elasticityof demand for x1, which I assume is negative. Note that η(τC) is theelasticity of demand with respect to the perceived price p + τC(θ );I denote by η(τ ) the elasticity of demand with respect to the actualprice p + τ .

To see how consumer responsiveness to the tax changeswith the salience of the tax, I will estimate empirically how theelasticity of demand with respect to the actual price (η(τ )) varieswith the tax salience (θ ). The sign of this relationship (i.e., the signof ∂η(τ )/∂θ) is ambiguous. To see this, note that the relationshipbetween η(τ ) (which I will estimate empirically) and η(τC) (whichI have assumed is constant) can be derived as follows:

η(τ ) ≡ ∂x1

∂(p + τ )(p + τ )

x1= ∂x1

∂(p + τC)∂(p + τC)∂(p + τ )

(p + τ )x1

p + τC

p + τC

= η(τC)p + τ

p + τC

∂(p + τC)∂(p + τ )

.(3)

Under the assumption of fixed producer prices (i.e., p does notvary with either τ or θ ), the relationship between the perceivedtax and actual tax in equation (1) implies that

(4)∂(p + τ )∂(p + τ )

= ∂τ

δτ= δ1c(θ ).

1. The assumption of quasi-linear utility seems a reasonable one whenthe taxed good is a small part of the overall consumer’s budget (such as the toll caseI consider). It is not, however, an innocuous assumption for the political responseto tax salience; I discuss this in more detail in Section II.B.



Using (4), we can simplify the relationship between η(τ ) and η(τC)in (3) to

(5) η(τ ) = η(τC)(

p + τ

p + τC

)δ1C(θ ).

Differentiating both sides of (5) with respect to salience (θ ) gives

∂η(τ )∂θ

= η(τC)(p + τ )︸︷︷︸−

×

⎛⎜⎜⎝ −1

(p + τC)2︸︷︷︸−

(δ′

0C(θ ) + δ′1C(θ )τ

)︸︷︷︸

?

δ1C(θ︸︷︷︸+

) + 1(p + τC)︸︷︷︸

+

δ′1C(θ︸︷︷︸−

)

⎞⎟⎟⎠ .(6)

Equation (6) shows that the sign of the impact of tax salience onthe elasticity of demand (i.e., the sign of ∂η(τ )/∂θ) is ambiguous,because the impact of salience on the level of the perceived tax(i.e., ∂τC/∂θ ≡ (δ′

0C(θ ) + δ′1C(θ )τ )) is of ambiguous sign.2 In the em-

pirical work I find evidence that consumption behavior becomesless elastic as salience decreases (i.e., ∂η(τ )/∂θ > 0). Equation (6)indicates that a sufficient (although not necessary) condition for∂η(τ )/∂θ > 0 is that δ′

0C(θ ) + δ′1C(θ )τ > 0 (i.e., the perceived tax is

increasing as salience decreases). In Section III I present surveyevidence that is consistent with this condition, suggesting thatthese empirical findings are internally consistent.

To estimate ∂η(τ )/∂θ empirically, I multiply (5) through by∂ log(p + τ ) to obtain

(7) ∂ log x1 = η(τC)(

p + τ

p + τC

)δ1C(θ )∂ log(p + τ ).

Taking a linear approximation to (7) around θ = 0 and explicitlyseparating out the main effects from the interaction effect of in-terest, I estimate

(8) � log(x1) = β1� log(p + τ ) + β2θ + β3θ� log(p + τ ) + �ε.

The parameter β1 provides an estimate of the estimated elasticityof demand in a fully salient system (i.e., θ = 0), in which case

2. The other components of (6) are signed by the assumptions discussed earlierin this section.



η(τC) = η(τ ) = β1. The parameter of interest is β3; it indicates howthe elasticity changes with salience.

II.B. Political Response to Tax Salience

The political response of tax rates to tax salience may dependnot only on how the consumer’s behavioral responsiveness to taxchanges with salience (i.e., ∂η(τ )/∂θ) but also on how the politicalcosts of taxes change with tax salience. Section II.A showed thatthe sign of the effect of tax salience on the consumer’s behavioralresponsiveness is ambiguous. Moreover, any effect of tax salienceon political costs need not be the same sign as any effect of taxsalience on consumer behavioral responsiveness, because salienceat the time of consumption and salience at the time of votingmay be different; this creates further ambiguity in the sign of therelationship between tax salience and tax rates. This ambiguitymotivates the empirical work that is the focus of this paper.

To gain some intuition into the determinants of the sign ofthe relationship between tax salience and tax rates, I consider agovernment that sets the tax to maximize a weighted sum of someeconomic objective and the (negative of) any political costs of thetax. For concreteness, I assume the economic objective of the tax isto raise revenue. I discuss other possible economic objectives—andhow these affect the implications of tax salience—in Section II.D.

The government chooses τ each year to maximize

(9) maxτ

λτ Q(p + τC) − (1 − λ) f (E) C(τv),

where 0 ≤ λ ≤ 1 represents the weight the government places onthe economic objective of the tax (i.e., raising revenue) relative tothe political cost of the tax, C denotes the political cost of the tax,and E is an indicator variable for whether or not it is an electionyear. I assume that f (E) > 0 and f ′(E) > 0; in other words, thepolitical costs of taxes are exogenously higher in election years,so that we expect a “political business cycle” in taxes (Nordhaus1975); in the empirical work, I provide evidence of a political busi-ness cycle in toll setting.

The government’s optimization problem yields the first-ordercondition for the tax rate

(10) τ ∗ = −Q(τC)Q′(τC)

+ (1 − λ) f (E) C ′(τV )λQ′(τC)

,



where to simplify notation I have defined C ′ ≡ (∂C/∂τ ) (∂τ/∂τ ) andQ′ ≡ (∂Q/∂τ ) (∂τ/∂τ ). To ensure an interior solution to the optimaltax, I assume that C ′ > 0 (i.e., political costs are rising in theactual tax) and Q′ < 0 (i.e., demand is falling in the actual tax).Note that both consumption salience and voting salience affectthe choice of tax rate: the amount of revenue raised depends onthe perceived tax at the time of the consumption decision (i.e., τC),and the political cost of the tax depends on the perceived tax atthe time of voting (i.e., τv).

Differentiation with respect to θ of the first-order conditionfor the government’s optimal tax level in (10) indicates that thesign of any effect of tax salience on the choice of tax rate is a prioriambiguous:

(11)∂τ ∗

∂θ=

⎛⎜⎜⎜⎜⎜⎜⎝

∂

(− Q

Q′

)

∂θ︸︷︷︸?

+ (1 − λ) f (E)λ︸︷︷︸+

⎛⎜⎝

∂C ′

∂θQ′ − ∂Q′

∂θC ′

(Q′)2

⎞⎟⎠

︸︷︷︸?

⎞⎟⎟⎟⎟⎟⎟⎠

.

Although the sign of (11) is theoretically ambiguous, thereare intuitive findings concerning how the relationship betweentax salience and tax rates is likely affected by the effect of salienceon the consumer’s behavioral responsiveness to taxes, and by theeffect of salience on the political costs of taxes. To see this, considerfirst the simplest case in which λ = 1, so that the government onlymaximizes revenue. In that case, the politically optimal tax inequation (10) reduces to the standard inverse elasticity optimaltax equation

(12)τ ∗

p + τ ∗ = 1η(τ )

,

and thus (under the assumption of fixed producer prices)

(13) sign of∂τ ∗

∂θ= sign of

1η(τ )2︸︷︷︸

+

∂η(τC)∂θ︸︷︷︸?

.

Equation (13) indicates that, when the government sets taxes tomaximize revenue, the sign of how taxes vary with salience isthe sign of how the elasticity of demand with respect to the taxvaries with salience (which as we saw in (6) can be of either sign).Intuitively, if a decline in salience lowers the behavioral response



to the tax (i.e., ∂η(τ )/∂θ > 0), then the tax rate set by the govern-ment will be rising as salience declines. Note that the assumptionof quasi-linear utility is important for this result, as it removesany distortionary effect of reduced salience on consumption of thetaxed good that arises from the budgetary consequences of themisperceived tax. In the more general case, where such distor-tionary effects will exist, Chetty, Kroft, and Looney (forthcom-ing) show that even if reduced salience reduces the behavioralresponse to the tax, this is not sufficient for the optimal tax toincrease; this is likely to be particularly important for taxes thatare a large share of the individual’s budget, such as income taxes.

Moreover, if the government puts some weight on the politicalcosts of taxes (i.e., λ < 1), this introduces another source of inde-terminacy in the sign of the relationship between tax salienceand tax rates. However, the model suggests that we can learnmore about the likely sign of ∂τ ∗/δθ in (11) by examining howany political business cycle in tax setting changes as tax saliencedeclines. To see this, note that

(14)∂2τ ∗

∂θ∂E=

(1 − λ

λ

f ′(E))

︸︷︷︸+

⎛⎜⎝

∂C ′

∂θQ′ − ∂Q′

∂θC ′

(Q′)2

⎞⎟⎠

︸︷︷︸?

and observe that the first term in parentheses is positive by as-sumption, and that the second term in parentheses (whose signis unknown) also appears in (11). Thus if ∂2τ/∂θ∂E > 0, this im-plies that the second term in parentheses in (14) is positive, sothat the entire second term in (11) is positive. In other words, ifthe political business cycle attenuates as salience declines (i.e.,∂2τ/∂θ∂E > 0, for which I find evidence in the empirical work be-low), this makes it more likely that a decline in tax salience raisestaxes (i.e., ∂τ ∗/δθ > 0).

To investigate the relationship between tax salience and taxrates empirically, I note that the first-order condition for the taxrate in (10) indicates that the tax rate will depend on tax salience(θ ), whether it is an election year (i.e., E = 1 or E = 0), and theinteraction of these two effects. Because of the serial correlationproperties of taxes in my empirical application (which I discuss inmore detail below), I estimate the relationship between taxes andsalience in first differences, estimating that

(15) �τ = β1�θ + β2 E + β3 E(�θ ) + �μ.



Estimation of (15) allows a comparison of the effect of tax salienceon tax rates in nonelection years (i.e., β1) and in election years(i.e., β3).

II.C. Identification

An examination of the two main estimating equations—equation (8), which comes from the driver optimization problemand reveals how behavioral responsiveness to the tax changeswith salience, and equation (15), which comes from the politi-cal optimization problem and reveals how the tax varies withsalience—highlights two important identification problems. First,taxes are taken as exogenous to demand in the demand estima-tion equation (8), but are determined as the endogenous result ofthe political optimization problem (see (10)). Identification of thedemand equation requires that the error term �ε in the demandequation (8) be uncorrelated with the error term �μ in the tax-setting equation (15); in other words, identification requires thatchanges in demand do not contemporaneously affect changes intaxes. For example, if demand follows a random walk, then as longas the government tax-setting process takes at least one year torespond to demand, current changes in taxes will be uncorrelatedwith current changes in demand and the demand equation (8) willbe identified.3

This identifying assumption seems reasonable for a (bureau-cratic) government that may not be able to make and implementdecisions quickly. In the empirical application, I will show that,in practice, taxes are changed only about once a decade, whichis consistent with the assumption of a lagged response. Further-more, any changes in taxes that are driven by changes in any ofthe nondemand factors that (10) indicates affect tax rates—thatis, the sensitivity of political costs to the tax rate (C ′), the elec-toral calendar (E), or the relative weight (λ) that the governmentplaces on the political costs of taxes—do not pose a problem foridentification (as long as changes in these factors are themselvesexogenous to changes in current demand).

The second identification problem is that I allow the tax (τ )to be chosen endogenously by the political optimization prob-lem in (9), but assume that the salience of the tax system (θ ) is

3. In my empirical application I find that changes in (residual) demand havean AR1 coefficient of 0.045, suggesting that demand is (close to) a random walk.I also explore robustness of demand estimation to alternative specifications withweaker identifying assumptions (see Section V).



exogenously determined. If the government endogenously choosesθ (e.g., on the basis of any of the factors that determine τ ), the tax-setting estimating equation (15) is not identified. The validity ofthe assumption that the choice of tax salience is exogenous withrespect to the choice of tax rate is ultimately an empirical ques-tion, and one that I explore in depth in Section VI.A.

II.D. Other Government Objective Functionsand Normative Implications

For concreteness, in Section II.B I assumed the government’sobjective function in choosing the tax rate was a weighted averageof the revenue raised by the tax (its economic objective) and the(negative of) the political costs of the tax (its political objective). Ofcourse, the government may well have other economic objectives,such as redistributive taxes or Pigouvian corrective taxes; the lat-ter is potentially quite relevant for the toll case that is the subjectof the empirical work. As with a revenue-raising tax, the optimallevel of these other types of taxes also varies inversely with thebehavioral responsiveness to the tax. For example, if the tax isset as an optimal Pigouvian externality correction, the optimaltax will be increasing as the behavioral responsiveness to the taxdeclines. Therefore the same empirical prediction concerning howthe impact of salience on the behavioral responsiveness to the taxlikely affects the impact of tax salience on tax rates should apply(qualitatively) to these other economic objectives.

In contrast to the positive empirical predictions, the norma-tive implications of any effect of tax salience on tax rates will bequite sensitive to the government’s objective function. One criticalissue for the normative implications of tax salience is whether thegovernment operates as a benign social planner or is (partially orfully) maximizing independent objectives (such as keeping politi-cians in office or increasing the size of government); in the lattercase, the government’s response to a decline in salience may beself-serving, but not socially optimal. The evidence I present be-low that the political business cycle in toll setting attenuates whensalience is reduced suggests that part of the impact of tax salienceon tax rates comes from reducing the political costs of raising tolls;this suggests that the government’s response to a reduction in taxsalience may not be that of a fully benign social planner.

Even when the government operates as a fully benign so-cial planner, the normative implications of a decline in saliencewill also depend on the economic component of the government’s



objective function. If the economic objective is to raise revenue,then if salience reduces the behavioral responsiveness to the tax,this is likely to be welfare-improving because it allows the govern-ment to raise a given amount of revenue at lower distortionarycosts. However, if the economic objective of the tax is a Pigou-vian externality correction, the normative implications may bequite different. For example, if salience reduces the behavioralresponsiveness to the tax, this has no effect on welfare if the taxis set solely as a Pigouvian corrective tax, utility is quasi-linearin the taxed good, and the revenue raised is rebated back to con-sumers as a lump sum; the government would raise the tax to the(new) higher optimal externality-correction tax and rebate backthe resulting (higher) revenue as a lump sum, with no changein aggregate welfare. However, in more general models in whichutility is not quasi-linear and/or the government does not rebateback the revenue raised as a lump sum, a lower behavioral re-sponsiveness to the Pigouvian tax due to reduced salience can bewelfare-reducing.

III. IMPACT OF ETC ON TOLL SALIENCE: SURVEY EVIDENCE

The empirical analysis is predicated on the assumption thatETC reduces the salience of the tolls (i.e., increases θ ). I there-fore begin by presenting survey evidence consistent with thisassumption.

Evidence from two separate surveys indicates that individu-als are substantially less aware of tolls if they pay them electron-ically rather than with cash. One survey is an in-person surveythat I designed and conducted in May 2007 of 214 individualswho had driven to an antiques show in western Massachusettson the Massachusetts Turnpike (“MA Survey”). The other is atelephone survey conducted in June and July 2004 of 362 regu-lar users from New Jersey of any of the six bridges or tunnels ofthe Port Authority of New York and New Jersey that cross theHudson River (“NYNJ Survey”). More details on the MA Surveycan be found in the Online Appendix (Section A); more details onthe NYNJ Survey can be found in Holguin-Veras, Kaan, and deCerrano (2005, especially pp. 116–126 and pp. 383–394).

Each survey asked drivers their estimate of the toll paid ontheir most recent trip on the relevant facility, their method ofpayment, and a variety of demographic characteristics; informa-tion about the exact trip was also collected so that the actual tollpaid could be calculated.



Table I summarizes the results. Both surveys show a strik-ingly lower awareness of tolls among drivers who paid with ETCthan among those who paid with cash. The differences are botheconomically and statistically significant. In the MA survey, 62%of drivers who paid using ETC responded to the question abouttheir best guess of the toll they paid that day on the Turnpikewith “I don’t know” and would not offer a guess without prompt-ing from the surveyor to please “just make your best guess”;4

in contrast, only 2% of drivers who paid with cash had to beprompted to offer a guess. In the NYNJ survey, 38.1% of ETCusers reported “do not know” or “refused” when asked how muchthey paid at the toll in their most recent drive across the Hud-son from New Jersey to New York, compared to 20.0% of cashusers.5

Moreover, the ETC drivers’ belief that they did not know howmuch they had paid for the toll was borne out by their subsequentguesses. In the MA Survey, 85% of drivers who paid using ETCestimated the toll they paid incorrectly, compared to only 31% ofdrivers who paid using cash. In the NYNJ survey, 83% of ETCdrivers estimated the toll incorrectly, compared to only 40% ofcash drivers. Conditional on making an error, the magnitude ofthe error was also larger for ETC users; ETC users overestimatetolls by more than cash users.6

These findings of markedly lower knowledge of tolls amongpeople who paid electronically than among those who paid withcash are consistent with the maintained assumption that tolls areless salient under ETC. In other words, the results are consistentwith ETC reducing the link between the actual and the perceivedtoll (i.e., δ′

1 j(θ ) < 0). These findings are also consistent with otherwork on “payment decoupling,” which finds that technologies suchas credit cards, which decouple the purchase from the payment,reduce awareness of the amount spent and thereby encouragemore spending (e.g., Thaler [1999]; Soman [2001]).

4. Indeed, many of the ETC drivers literally responded, “I don’t know, I usedEZ-Pass [or Fast Lane].”

5. It is interesting that the discrepancy in toll awareness between ETC andcash drivers is larger in the MA survey. One possible explanation is that theNYNJ Survey asked about the toll paid on a regular commute, whereas the MASurvey asked about the toll paid on a presumably idiosyncratic trip. Differencesin the survey method (e.g., telephone vs. in person) may also have an effect on theindividual’s willingness to guess.

6. This finding that ETC is associated with overestimation of the toll is con-sistent with the finding in Section V that ETC is also associated with reducedbehavioral responsiveness to the toll. See equation (6) in Section II.A and thediscussion that follows it.


982 QUARTERLY JOURNAL OF ECONOMICST

AB

LE

IS

UR

VE

YE

VID

EN

CE

ON

DR

IVE

RA

WA

RE

NE

SS

OF

TO

LL

S,B

YP

AY

ME

NT

ME

TH

OD

MA

surv

eyN

YN

Jsu

rvey

Dif

fere

nce

betw

een

ET

Can

dca

shdr

iver

sD

iffe

ren

cebe

twee

nE

TC

Cas

hN

oC

ovar

iate

ET

CC

ash

ET

Can

dca

shdr

iver

sdr

iver

sdr

iver

sco

vari

ates

adju

sted

driv

ers

driv

ers

(no

cova

riat

es)

(1)

(2)

(3)

(4)

(5)

(6)

(7)

Fra

ctio

nw

ho

repo

rt“d

on’t

know

”0.

618

0.02

10.

597∗

∗∗0.

579∗

∗∗0.

381

0.20

00.

18∗∗

∗(0

.490

)(0

.142

)(0

.060

)(0

.060

)(0

.486

)(0

.400

)(0

.05)

Fra

ctio

nw

ho

inco

rrec

tly

0.85

10.

308

0.54

3∗∗∗

0.51

2∗∗∗

0.82

60.

395

0.43

∗∗∗

esti

mat

eto

ll(0

.359

)(0

.463

)(0

.058

)(0

.067

)(0

.379

)(0

.489

)(0

.06)

Mea

ner

ror,

con

diti

onal

$1.3

34$0

.162

$1.1

72∗∗

∗$1

.01∗

∗∗$0

.40

−$0.

10$0

.50

onm

isre

port

ing

(1.8

50)

(0.8

28)

(0.2

75)

(0.3

03)

N68

146

271

91

Not

es.I

nco

lum

ns

(1),

(2),

(5),

and

(6),

stan

dard

devi

atio

ns

are

inpa

ren

thes

es;i

nco

lum

ns

(3),

(4),

and

(7)r

obu

stst

anda

rder

rors

are

inpa

ren

thes

esan

d∗∗

∗ ,∗∗

,∗de

not

est

atis

tica

lsi

gnifi

can

ceat

the

1%,5

%,a

nd

10%

leve

ls,r

espe

ctiv

ely.

“Err

or”

inth

eth

ird

row

isco

mpu

ted

ases

tim

ated

toll

−ac

tual

toll

paid

.In

the

MA

Su

rvey

anes

tim

ate

ofth

eto

llpa

idw

asev

entu

ally

elic

ited

from

all

but

one

ofth

ere

spon

den

ts;

how

ever

,in

the

NJN

YS

urv

ey,

anes

tim

ate

ofth

eto

llpa

idw

ason

lyel

icit

edfo

rth

ose

wh

odi

dn

otre

spon

d“d

on’t

know

”or

“ref

use

d.”

Th

us,

for

the

MA

Su

rvey

,th

esa

mpl

ein

row

s(2

)an

d(3

)in

clu

des

allb

ut

one

ofth

ere

spon

den

tsin

row

(1),

but

for

the

NY

NJ

Su

rvey

,th

esa

mpl

ein

row

s(2

)an

d(3

)in

clu

des

only

thos

ere

spon

den

tsw

ho

did

not

repo

rt“d

on’t

know

”in

row

(1).

For

the

NY

NJ

surv

ey,

the

cash

toll

was

$6.0

0,w

her

eas

the

ET

Cto

llw

as$5

.00

onpe

akan

d$4

.00

off

peak

.F

orth

eM

Asu

rvey

,th

eto

llde

pen

ded

onth

een

tran

cean

dex

itta

ken

.T

he

aver

age

toll

paid

was

abou

t$1

.15.

Les

sth

an10

%of

driv

ers

inth

eM

Asu

rvey

sam

ple

drov

eon

apo

rtio

nof

the

Tu

rnpi

kein

wh

ich

ther

ear

eE

TC

disc

oun

ts,

and

the

resu

lts

are

not

affe

cted

byom

itti

ng

thes

edr

iver

sfr

omth

ean

alys

is.

Inco

lum

n(4

),co

vari

ates

con

sist

ofag

e,ag

esq

uar

ed,

med

ian

hou

seh

old

inco

me

ofZ

IPco

de,

deal

erre

tail

pric

efo

rth

edr

iver

’sca

r(b

ased

onin

form

atio

nfr

omw

ww

.edm

un

ds.c

omas

ofO

ctob

er20

07),

and

indi

cato

rva

riab

les

for

sex,

wh

eth

erth

edr

iver

regu

larl

ypa

ysa

toll

ona

com

mu

teto

wor

k,an

dh

igh

est

leve

lof

edu

cati

onre

ach

ed(h

igh

sch

oold

egre

eor

less

,col

lege

degr

ee,o

rpo

stco

lleg

ede

gree

,wh

ere

“col

lege

degr

ee”

incl

ude

sas

soci

ates

degr

ees,

wh

ich

wer

e10

%of

the

coll

ege

degr

eesa

mpl

e).O

nly

publ

ish

edsu

mm

ary

stat

isti

cs(a

sop

pose

dto

the

un

derl

yin

gm

icro

data

)ar

eav

aila

ble

for

the

NY

NJ

surv

ey,

soth

atth

eco

vari

ate-

adju

sted

diff

eren

cein

mea

ns

can

not

beco

mpu

ted.

Inad

diti

on,

the

sam

ple

size

sby

cell

for

the

NY

NJ

surv

eyh

adto

beap

prox

imat

edba

sed

onin

form

atio

nin

the

text

onth

eto

tals

ampl

esi

ze(3

62)a

nd

the

frac

tion

ofdr

iver

sth

atpa

yby

ET

C(7

4.8%

).A

sa

resu

lt,t

he

stan

dard

erro

rsfo

rth

eN

YN

JS

urv

eyar

eal

soap

prox

imat

ed;

appr

oxim

ated

nu

mbe

rsar

esh

own

init

alic

s.I

calc

ula

ted

stan

dard

devi

atio

ns

for

the

bin

ary

resp

onse

vari

able

sin

the

NY

NJ

Su

rvey

,bu

tth

ere

was

not

suffi

cien

tin

form

atio

nav

aila

ble

toca

lcu

late

the

stan

dard

devi

atio

nfo

rth

em

ean

erro

r(o

rth

est

anda

rder

ror

ofth

edi

ffer

ence

inm

ean

erro

r).



Several caveats are in order. First, neither survey is represen-tative of the nationwide population. Nonetheless, it is reassuringthat the finding of lower toll awareness among ETC drivers per-sists in two very different populations, including a population ofregular commuters. Second, cross-sectional differences in aware-ness of tolls between ETC drivers and cash drivers could reflectdifferences in these drivers besides their payment method. Re-assuringly, a comparison of the results in columns (3) and (4) ofTable I shows that none of the differences in toll awareness inthe MA Survey are sensitive (in either magnitude or statisticalsignificance) to adding controls for demographic characteristics ofdrivers, including age, sex, education, median household incomeof ZIP code, and value of their car.

Finally, a survey response on toll perception does not neces-sarily reflect either the perceived toll at the time of consumption(τC) or the perceived toll at the time of voting (τV ). However, giventhe large percentage of cash drivers relative to ETC drivers whoare spot on in estimating the toll paid correctly, it seems plausiblethat ETC may reduce one or both of these types of salience. I nowturn to direct evidence of the impact of ETC first on consumerbehavior and then on toll setting.

IV. DATA AND DESCRIPTIVE STATISTICS

This section provides some brief background on the sampleconstruction and variable definitions for the toll facility data; con-siderably more details on the facilities in the sample and the vari-able definitions can be found in the Online Appendix (Section B)or in the working paper version of this paper (Finkelstein 2007).

IV.A. Sample Construction

The target sample was all 183 publicly owned toll facilitiesin the United States (excluding ferries) that were charging tollsin 1985, which predates the introduction of ETC in the UnitedStates. In 1985, toll revenue in states that levied tolls was about0.8% of state and local tax revenue, roughly the same revenueshare as state lotteries (U.S. Census Bureau 1985; U.S. Depart-ment of Transportation 1985, 1986; Kearney 2005). Statutoryauthority for toll setting is usually vested in toll operating author-ities. These are typically appointed by state or local governments,which therefore, in practice, retain influence on toll setting.


984 QUARTERLY JOURNAL OF ECONOMICS0

51

01

5F

req

ue

ncy

1985 1990 1995 2000 2005ETC start date

FIGURE IDistribution of ETC Start Dates

By contacting each toll authority, I was able to collect data for123 toll facilities.7 These 123 facilities are run by 49 different op-erating authorities in 24 different statelike entities; these include22 states and 2 joint ventures (one between New York and NewJersey and one between New Jersey and Pensylvania). I refer toall 24 hereafter as “states.” On average, the data contain 50 yearsof toll rates per facility.

IV.B. Key Variables

ETC Adoption and Diffusion. Figure I shows a histogram ofETC adoption dates, which range from 1987 through 2005, with amedian of 1999. By 2005, 87 of the 123 facilities had adopted ETC.Almost all of the variation in whether and when ETC is adopted isbetween rather than within operating authorities; there is, how-ever, substantial variation across authorities within a state (notshown). On average for a facility with ETC, I observe about sixyears of ETC.

Table II shows that relationship between facility character-istics and ETC adoption. ETC adoption rates are highest in thenortheast (78%) and lowest in the west (57%). The high adoptionrates in the northeast may reflect greater urbanism (because ETC

7. A toll “facility” is a particular road, bridge, or tunnel; about 60 percent ofthe responding facilities are bridges or tunnels.



TABLE IIWHICH FACILITIES ADOPT ETC?

Probability of Average adoptionNumber of adopting date conditionalfacilities ETC by 2005 on adoption

All 123 .71 1998.2By facility type

Roads 44 .70 1996.4Bridges or tunnels 79 .71 1999.2

By region of countryNortheast 58 .78 1998.7Midwest 10 .60 1996.7South 41 .68 1997West 14 .57 2000.9

may help reduce congestion) as well as higher labor costs (becauseETC reduces labor costs of toll collection). ETC is adopted withthe same probability on roads as on bridges and tunnels; however,roads that adopt ETC do so about three years earlier on averagethan bridges or tunnels that adopt ETC. Older facilities are morelikely to adopt ETC, and those that do are likely to do so earlierthan younger facilities that adopt ETC (not shown).

Once a facility adopts ETC, use of the technology diffusesgradually across drivers. I was able to obtain the ETC penetrationrate (defined consistently within each facility as either the fractionof toll transactions or the fraction of toll revenue collected by ETC)for about two-thirds of facility-years with ETC. Figure II shows thewithin-facility ETC diffusion rate. It takes about fourteen yearsfor ETC to reach its steady state penetration rate of 60 percent.

Toll Histories. I define the toll as the nominal toll for passen-ger cars on a full-length trip on a road, or on a round trip on abridge or tunnel. I collected data on both the “manual” (i.e., cash)toll and any discount offered for the electronic toll; the electronictoll is never more than the cash toll.8 Over half (53 of 87) of fa-cilities with ETC offer a discount at some point. Discounts arepresumably offered to encourage use of the technology; indeed,they are more common on facilities that adopt ETC earlier. Thediscounts may also be rationalized as a Pigouvian subsidy if ETChas positive externalities on congestion reduction. The averagediscount offered is about 15 percent.

8. High-frequency discounts (i.e., commuter discounts) are not coded. None ofthe facilities in the sample offer time-of-day varying prices.



0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

ETC year

ET

C p

en

etr

ati

on

ra

te

FIGURE IIWithin-Facility ETC Diffusion

Figure II reports the coefficients on indicator variables for the number of yearsa facility has had ETC from the following regression: ETC Penetrationit = αi +∑19

k=1 βk1(ETCyear = k), where the αi are facility fixed effects, 1(ETCyear = k) areindicator variables for whether it is the kth year of ETC, and ETC Penetration isdefined either as percentage of toll transactions paid by ETC or as percentage ofrevenue paid by ETC, depending on the facility. The regression is estimated onthe sample of facility-years with ETC and data on ETC penetration (N = 467; 84unique facilities).

The primary toll measure in the analysis is the lower envelopeof the manual and electronic tolls (hereafter, “minimum toll”).I also present results for the subsample of facilities that neveroffer ETC discounts, and for which the minimum and manualtoll are therefore always the same. On average, the minimum tollincreased by 2.0% per year. This is substantially below the facility-year-weighted average inflation rate of 4.2%. Toll changes arelumpy; on average only 7.7% of facilities increase their minimumtoll and only 1% of facilities decrease it each year.

Revenue and Traffic Data. I was able to collect traffic (rev-enue) data for 76 (45) of the 123 facilities. On average, for a facilitywith these data, I obtained 34 years of data.

V. THE IMPACT OF ETC ON THE ELASTICITY OF DRIVING

WITH RESPECT TO THE TOLL CHANGE

To examine how ETC affects the elasticity of driving withrespect to the tax, I adapt the demand equation (8) to the toll



context as follows:

� log(traffic)it = γt + β1� log (minimum tollit)

+β2� log (minimum tollit) ∗ Never ETCi

+β3� log (minimum tollit) ∗ ETC Penetrationit

+β4 Never ETCi + β5 ETC Penetrationit + �εit.(16)

I proxy for demand for the taxed good (i.e., x1 in (8)) with theamount of traffic on facility i in year t (i.e., trafficit), and for thesalience of the tax system (i.e., θ in (8)) with the ETC Penetrationrate on facility i in year t (i.e., ETC Penetrationit). For purposesof practicality, I estimate the demand responsiveness to τ in (16)rather than to p + τ as in (8), because I do not observe the non-tax costs (p) of driving. As long as p does not vary with taxes orwith tax salience (i.e., the fixed producer prices assumption dis-cussed in Section II), this modification will affect the magnitudeof the estimated elasticities but not their sign. As noted, I use theminimum toll as my measure of τ .

Equation (16) examines the relationship between the annualpercentage change in a facility’s traffic (�log(traffic)it) and theannual percentage change in its toll (�log(minimum toll)it) andhow this relationship changes with the ETC penetration rate.To strengthen the inference, it also allows the elasticity to varyacross facilities based on whether the facility ever adopted ETC(Never ETCi is 1 if the facility never adopts ETC and zero other-wise), and it allows for secular changes in demand over time (theγt represent a full set of year fixed effects). The key coefficient ofinterest is β3; this indicates how the elasticity changes at a facilityas ETC use diffuses. Finally, �εit is a random disturbance termcapturing all omitted influences. I allow for an arbitrary variance–covariance matrix within each “state” and give equal weight in theregression to each operating authority.

As discussed in Section II.C, identification of (16) is basedon the assumption that changes in tolls are not affected bycontemporary changes in demand. This is probably a reason-able assumption. Traffic—and presumably underlying demand fordriving—changes continuously each year, whereas a facility’s tollis raised on average only every eight to nine years. The infre-quency of toll adjustment likely reflects both general lags in pricesetting by government enterprises and political constraints; forexample, I show in Section VI.B that toll increases are signifi-cantly lower during state election years. Although tolls may be



adjusted in part based on past demand shocks (i.e., lagged valuesof changes in traffic), changes in traffic within a facility show verylittle serial correlation; a regression of the residuals from (16) ontheir lags produces a coefficient of only 0.045. Any adjustment oftolls to past changes in demand is therefore unlikely to pose muchof a practical problem for the estimation. However, as a robust-ness check, I also report results in which I limit the sample to theyears in which a toll changes or the two years before or after a tollchange; I refer to this as the “+2/−2 sample.” The assumption inthis more limited sample is that the timing of the toll change israndom with respect to short-run traffic changes, although it mayreflect longer-run demand changes.

I estimate (16) on approximately one-fourth of the facilitiesin the data. By necessity, the analysis is limited to the approx-imately 60 percent of facilities for which I obtained traffic data.I further limit the subsample of facilities with traffic data to theapproximately 40 percent of them that never offer an ETC dis-count. This allows me to include the ETC penetration rate directlyon the right-hand side, without worrying about omitted variablebias from any potential effect of an ETC discount on both the ETCpenetration rate and traffic. An added advantage of looking onlyat facilities that never offer an ETC discount is that in this samplethere is only one toll rate (i.e., the minimum toll and the toll arealways the same), which avoids the measurement error that ETCdiscounts would otherwise introduce in the right-hand-side tollvariable once ETC is introduced.9

Table III reports the results. Columns (1) and (2) show the re-sults from regressing �log(traffic)it on �log(minimum toll)it andyear fixed effects. Column (1) shows the results for the full sam-ple of facilities with traffic data, including those that offer ETCdiscounts. The coefficient on �log(minimum toll)it of −0.049 (stan-dard error 0.015) indicates that a 10% increase in tolls is associ-ated with a statistically significant but economically small 0.5%reduction in traffic. Column (2) shows that the result is quite sim-ilar for the sample of facilities that never offer ETC discounts;the coefficient on �log(minimum toll)it is −0.058 (standard error

9. I show below that the estimated impact of ETC on toll rates is robust tolimiting the sample to facilities that never offer discounts. When I limit to those forwhom I have traffic data, the effect is very similar in magnitude to the estimates inthe full sample, although no longer statistically significant at conventional levels(not shown).



TABLE IIITHE ELASTICITY OF TRAFFIC WITH RESPECT TO TOLLS

(1) (2) (3) (4) (5) (6)

� log min. tollit −0.049 −0.058 −0.061 −0.057 −0.062 −0.060(0.015) (0.018) (0.019) (0.017) (0.039) (0.037)

[.004] [.008] [.009] [.006] [.145] [.135]� log min. tollit * 0.134 0.141

ETC penetrationit (0.038) (0.076)[.005] [.091]

� log min. tollit * 0.006 0.006ETC yearit (0.001) (0.003)

[.002] [.062]� log min. tollit * −0.071 −0.073 −0.009 −0.006

never ETCi (0.136) (0.131) (0.209) (0.205)[.611] [.588] [.966] [.976]

Mean dep. var. 0.049 0.042 0.043 0.042 0.040 0.039# of states 21 12 12 12 12 12# op. authorities 32 16 16 16 16 16# of facilities 76 33 33 33 33 33N 2,200 727 671 727 292 305Sample No ETC No ETC No ETC No ETC No ETC

restriction(s) discounts discounts discounts discounts discounts+2/−2 +2/−2sample sample

Notes. Table reports results from estimating variants of (16) by OLS. The dependent variable is the changein log traffic. In addition to the covariates reported in the table, all regressions include year fixed effects anda main effect for any variables that are interacted with � log(min. toll). The bottom row indicates any samplerestrictions. “No ETC discounts” limits facilities to those that never offered an ETC discount. “+2/−2 sample”limits sample to facility-years in which there is a toll change or the two years before or after a facility’s tollchange. Never ETCi is an indicator variable for whether facility i never has ETC. ETC penetrationit is theshare of tolls paid by ETC on facility i in year t; it is zero in years in which the facility did not have ETC.ETC yearit is the number of years the facility has had ETC; it is zero in any year in which the facility doesnot have ETC, 1 the year the facility adopts ETC, 2 the second year the facility has ETC, and so forth. Eachoperating authority receives equal weight. Standard errors (in parentheses) are clustered by state. p-valuesare reported in square brackets.

= 0.018). These results suggest that tolls are set below the profit-maximizing rate, which is consistent with Peltzman’s (1971) ob-servation that there will be a downward bias in the prices set bygovernment-owned enterprises. More generally, it suggests that—as modeled in Section II.B—the government objective function isnot pure revenue maximization.10

10. Of course, I am only measuring the short-run response to a small change intolls; this behavioral response may merely reflect the route chosen on a particularday. Longer-run responses to (possibly larger) toll changes may be larger, reflectingamong other things decisions that affect regular commuting patterns.



Column (3) shows the results from estimating the com-plete equation (16). The coefficient on � log(minimum tollit) ∗ETC penetrationit is 0.134 (standard error 0.038); this indicatesthat a 5-percentage-point increase in the ETC penetration rate(which is the average increase per year of ETC) is associated witha (statistically significant) 0.0067 decline in the elasticity of driv-ing with respect to the toll, or about 10 percent relative to theaverage estimated elasticity prior to ETC of −0.061.

Column (4) shows the results when the ETC Penetration vari-able in (16) is replaced by the number of years the facility has hadETC (ETC Year); this variable is zero prior to ETC adoption, 1 inthe year of adoption, 2 in the second year of ETC, and so forth. Thecoefficient on � log(minimum tollit) ∗ ETC Yearit is 0.006 (stan-dard error 0.001), indicating a decline in elasticity of 0.006 peryear of ETC quite similar to that estimated in column (3).11

The last two columns of Table III repeat the analysis incolumns (3) and (4) on the +2/−2 sample. The point estimateson both the elasticity of driving under manual toll collectionand the change in the elasticity associated with ETC Year (orETC Penetration) remain virtually unchanged. The change in theelasticity associated with ETC remains statistically significant,although at the 10% level in the +2/−2 sample (columns (5) and(6)) rather than at the 1% level as in the larger samples (columns(3) and (4)).

As noted in Section II.B, for taxes that are small as a portion ofincome, if a decline in salience reduces the behavioral responsive-ness to the toll, this will tend to cause tolls to rise when saliencedeclines. However, the net impact of salience on toll rates is am-biguous; it also depends on how salience affects the political costsof toll setting. I now turn to an examination first of the net effectof ETC on toll rate and then of the effect of ETC on the politicalcosts of tolls.

11. One potential concern in interpreting these results is that the findingof a decline in the (absolute value) of the elasticity of driving with respect tothe toll under ETC might spuriously reflect a general time trend in the elas-ticity of driving with respect to the toll. To investigate this, I reestimated theregressions shown in columns (3) and (4) of Table III with the inclusion of anadditional interaction term �log(minimum toll)it * yeart on the right-hand side;this allows for a time trend in the elasticity of driving. The inclusion of this inter-action term weakened the precision of the estimated decline (in absolute value)of the driving elasticity under ETC, but did not substantively affect the find-ing. For example, for the specification shown in column (3), the coefficient on� log(minimum tollit) ∗ ETC penetrationit became 0.137 (standard error 0.067).In column (4), the coefficient on � log(minimum tollit) ∗ ETC Yearit became 0.005(standard error 0.002).



VI. THE IMPACT OF ETC ON POLITICAL BEHAVIOR

VI.A. The Impact of ETC on Toll Rates

Baseline Specification. To estimate the impact of ETC on tollrates, I begin with a simplified version of the estimating equa-tion for tax setting (equation (15)) in which I omit any measure ofwhether it is an election year from the right-hand side. Becausethe election calendar is set exogenously, this does not introduceany omitted variable bias, and allows me to capture the aver-age impact of ETC on toll rates; I augment the analysis to includeelectoral effects in Section VI.B.

I therefore begin with the estimating equation:

(17) �yit = γt + β1ETCAdoptit + β2ETCit + �μit.

In the baseline specification, the dependent variable is the changein the log of the minimum toll (�log(min toll)it). I estimate thedependent variable in logs rather than in levels (as in equation(15) in Section II.B) in order not to constrain toll rates in differentfacilities to grow by the same absolute amount each year; thisseems undesirable, given the considerable variation in toll ratesacross facilities.12 The γts represent year dummies that controlfor any common secular changes in toll rates across facilities.

The key coefficients of interest are those on ETCAdoptit andETCit, which represent my parameterization of the change in taxsalience (�θ in (15)). Specifically, ETCAdoptit is an indicator vari-able for whether facility i adopted ETC in year t. The coefficienton ETCAdoptit thus measures any level shift in the minimum tollassociated with the introduction of ETC; this might include, forexample, the effect of any ETC discounts. However, because ETCuse among drivers diffuses gradually, it is likely that any impactof ETC on toll rates will also phase in gradually. To capture this, Iinclude the indicator variable ETCit for whether facility i has ETCin year t; it is 1 in the year of ETC adoption and in all subsequent

12. In practice, the sign and statistical significance of the impact of ETC ontolls are robust to specifying the dependent variable as the change in the levelof the minimum toll rather than the change in the log of the minimum toll; themagnitude of the effect is slightly more than double in this alternative specification(not shown). One potential concern with the log specification is that the dependentvariable is censored when a toll is set to 0. Indeed, 15 of the 123 facilities thatwere charging a toll in 1985 subsequently set the toll to zero. I treat all facility-years with zero tolls as censored (both in the log and in the level analysis). Thislikely biases downward any estimated impact of ETC, because I find that ETC isassociated with a negative and marginally statistically significant decline in theprobability that the toll rate is changed from nonzero to zero (not shown).



years. The coefficient on ETCit thus measures the average annualgrowth in a facility’s toll once it has ETC. Thus I parameterize�θ with ETCAdoptit and ETCit in the first year of ETC, and Iparameterize �θ with ETCit in all subsequent years with ETC.

Finally, �μit is a random disturbance term capturing allomitted influences.13 I estimate (17), allowing for an arbitraryvariance–covariance matrix within each state, and give equalweight in the regression to each operating authority.

The first column of Table IV shows the results from estimat-ing (17). The coefficient on ETCit is 0.015 (standard error 0.006).This indicates that once a facility has ETC, its toll increases by1.5 percentage points more per year than it otherwise would have.This effect is both statistically and economically significant. Rel-ative to the average annual 2% increase in tolls, it implies thatafter installation of ETC, the facility’s toll rate rises by 75% moreper year than it did prior to ETC.14

The toll change in the first year of ETC is given by the sumof the coefficients on ETCAdoptit and ETCit. These indicate thatthere is a (statistically insignificant) 3.6% decline in tolls the yearthat ETC is adopted. The results in the next two columns sug-gest that this decline in the year of ETC adoption is due to ETCdiscounts. Column (2) shows the results when the dependent vari-able is the change in the log manual toll; column (3) shows theresults when the sample is limited to the approximately 60 per-cent of facilities that never offered an ETC discount (half of whichnever adopted ETC), for which the manual and minimum toll arealways the same. In these alternative specifications, the sum ofthe coefficients on ETCAdoptit and ETCit is either positive andinsignificant (column (2)) or negative and now both economicallyand statistically insignificant (column (3)).

The fact that the growth in tolls under ETC persists in the“no discount” sample (column (3))—the coefficient on ETCit is sta-tistically significant and slightly larger in magnitude than in thefull sample in column (1)—indicates that the estimated growth

13. I estimate (17) in first differences rather than in levels with facility fixedeffects because the residuals are much less highly serially correlated in first dif-ferences (AR1 coefficient of −0.045) than in the fixed effects version (AR1 coeffi-cient of 0.92), making the first-differenced specification the preferred specification(Wooldridge 2002, pp. 274–281).

14. One might prefer to specify the percentage increase in the toll associatedwith ETC relative to the average annual growth rate of tolls prior to ETC; this is1.9%. It is quite similar to the sample average (despite an average annual growthrate of tolls under ETC of 2.8%) because the vast majority of facility-years in theapproximately fifty-year toll histories I collected on each facility do not have ETC.



TABLE IVIMPACT OF ETC ON TOLL RATES

� log � log � log � log � log � logmin. toll manual toll toll toll min. toll min. toll

(1) (2) (3) (4) (5) (6)

ETCit 0.015 0.020 0.024(0.006) (0.006) (0.012)

[.018] [.004] [.061]�ETC 0.623 0.557 0.501

penetrationit (0.285) (0.262) (0.261)[.044] [.045] [.067]

ETCAdoptit −0.051 0.016 −0.033 −0.051 −0.105 −0.097(0.035) (0.032) (0.019) [0.035] (0.109) (0.108)

[.158] [.622] [.097] [.166] [.348] [.380]Mean dep. var. 0.020 0.022 0.017 0.017 0.020 0.020# of states 24 24 17 17 24 24# op. authorities 49 49 31 31 49 49# facilities 123 123 70 70 123 123N 5,079 5,079 2,875 2,751 4,815 4,815Estimation OLS OLS OLS OLS IV IVSample No ETC No ETC

restriction discount discount

Notes. Table reports results of estimating (17) (columns (1)–(3)) and (19) (columns (4)–(6)). Columnheadings define the dependent variable; the bottom two rows provide additional information on the estimationtechnique and sample restriction. ETCAdoptit is an indicator variable for whether facility i adopted ETC inyear t. ETCit is an indicator variable for whether the facility has ETC; it is 1 in the year that ETC is adoptedand in all subsequent years. �ETC penetrationit measures the change in the proportion of tolls on the facilitypaid by ETC; it is zero if the facility does not have ETC. In column (5), the instrument for �ETC penetrationitis ETCit . In column (6), the instrument for �ETC penetrationit is a cubic polynomial in the number of yearsthe facility has had ETC. In addition to the covariates shown in the table, all regressions include year fixedeffects. Each operating authority receives equal weight. Standard errors (in parentheses) are clustered bystate. p-values are reported in square brackets. “No ETC discounts” limits facilities to those that never offeredan ETC discount. Declines in sample size in column (4) (compared to column (3)) and in column (5) or (6)(compared to column (1)) reflect missing data on ETC penetration rates (see Section IV).

in tolls after ETC is installed does not merely reflect a recoupingof first-year losses from the ETC discount. For facilities that offerETC discounts, there does not appear to be any systematic changein the discount over time after ETC adoption (not shown). Thissuggests that in practice increases in the minimum toll reflecta shift of the entire toll schedule, which is consistent with thefinding that the manual toll also increases under ETC (column(2)).15

15. Although it might at first appear puzzling that the manual (i.e., cash)toll—which has become no less salient—also increases under ETC, this is easilyunderstood by the necessary linkage between cash and electronic toll rates; werethe electronic rate to increase while the cash rate did not, this would presumablydiscourage use of ETC. The preservation of the ETC discounts once ETC is installed



The Pattern of ETC Diffusion and Toll Increases. The preced-ing analysis constrains the effect of ETC to be the same acrossfacilities and over time. However, if ETC increases tolls by reduc-ing their salience, we would expect the effect to be increasing inthe ETC penetration rate, whose diffusion rate is not constant overtime (see Figure II) or across facilities (not shown). As a strongertest of the salience hypothesis, therefore, I examine how the timepattern of toll changes after ETC adoption compares to the timepattern of ETC diffusion. Specifically, I compare the coefficientsfrom estimating

(18a) � log(min toll)it = γt +k=9∑

k=−9

βk1(ETCYear(k,k+1)

) + εit

and

(18b) �ETC Penetrationit = γt +k=9∑k=1

βk1(ETCYear(k,k+1)

) + εit,

where �ETC Penetrationit is the percentage point change in theETC penetration rate for facility i in year t. The key outcome ofinterest is a comparison of the time pattern of the coefficients onthe indicator variables 1(ETCYear(k,k+1)) across the two equations.These are indicator variables for whether it is k or k + 1 years sinceETC was adopted on the facility. For example, 1(ETCYear(1,2)) is anindicator variable for whether ETC was adopted this year or lastyear (i.e., ETC Year is 1 or 2). In (18a), all of the indicator variablesrepresent a two-year interval, except for the first (respectively,last) indicator variable, which is a “catch-all” variable for whetherit is 9 or more years before (respectively, after) ETC adoption; theomitted category is the two years prior to adoption (i.e., ETC Yearof −1 or −2). In (18b) I include only the post-ETC dummies thatare in (18a).

Figure IIIA shows the result. The solid black line shows thepattern of the log toll with respect to ETC Year implied by theestimates from (18a) and the dark dashed line shows the corre-sponding time pattern of ETC diffusion implied by the estimates

likely reflects continued attempts to induce more drivers to switch to ETC; themaximum ETC penetration rate in my sample is only 78%.



–0.15

–0.1

–0.05

0

0.05

0.1

0.15

–8 –6 –4 –2 642 8

ETC year

–0.05

0.05

0.15

0.25

0.35

0.45

0.55

Log minimum toll (left axis)

ETC penetration rate (right axis)

(A)

–0.15

–0.1

–0.05

0

0.05

0.1

0.15

0.2

–8 –6 –4 –2 2 4 86

ETC year

–0.05

0.05

0.15

0.25

0.35

0.45

0.55

ETC penetration rate (right axis)

Log minimum toll (left axis)

(B)

FIGURE IIITime Pattern of Toll Changes and ETC Diffusion

The solid black line shows the pattern of log minimum toll implied by theestimates from (18a); the light dashed lines show the corresponding 95% confidenceinterval. The dark dashed line shows the pattern of the ETC penetration rateimplied by estimating (18b). ETC year represents the number of years since (orbefore) ETC adoption. The omitted category (ETC year −2 for (18a) and all yearsprior to ETC adoption for (18b)) is set to zero. Indicator variables for whether itis nine or more years after ETC adoption are included in the estimating equationbut not graphed; in (4a) an indicator variable for whether it is nine or more yearsbefore ETC adoption is also included in the regression but not graphed. In PanelB the sample of ETC-adopting facilities is limited to those who adopted in 1998or earlier. The upper end of the 95% confidence interval for the log minimum tollat eight years is not shown for scale reasons; it is 0.201 (full sample, A) and 0.311(balanced panel, B). To enhance the readability of the graph, the 95% confidenceinterval on ETC penetration rate is not shown. For Panel A the upper and lower95% confidence intervals for ETC penetration rate are as follows: (0.16, 0.378)for ETC year 2, (0.267, 0.484) for ETC year 4, (0.336, 0.565) for ETC year 6, and(0.378, 0.610) for ETC year 8. For Panel B, the analogous confidence intervals are(0.197, 0.283), (0.333, 0.425), (0.389, 0.550), and (0.419, 0.617).



of (18b).16 The results indicate that, after remaining roughly con-stant in the pre-ETC period, toll rates decline in the first two yearsof ETC (reflecting the discounts discussed earlier) and then climbsteadily as ETC diffuses across the facility. Of course, the wideconfidence intervals on the estimates caution against placing toomuch weight on the estimated time path. It is nonetheless reas-suring that the point estimates suggest that the pattern of tollincreases is similar to that of ETC diffusion.

A potential concern with this analysis is that the set of facili-ties that identify the different βks varies with the ETC year k. It istherefore difficult to distinguish the time path of the effect of ETCon a given facility from potentially heterogeneous effects of ETCacross facilities.17 Figure IIIB therefore shows the results fromre-estimating (18a) and (18b) when the sample of ETC-adoptingfacilities is limited to those that adopted ETC in 1998 or earlier.In this balanced panel of facilities, all of the graphed coefficientsare identified by a constant set of facilities. The results are quitesimilar.18

For a more parametric (and higher-powered) analysis of howthe time pattern of toll changes after ETC adoption compares withthe diffusion of ETC, I estimate a modified version of (17):

�log(min toll)it = γt +β1ETCAdoptit + β2�ETC Penetrationit + εit.

(19)

By replacing the indicator variable for whether the facilityhas ETC (ETCit) with the percentage point change in ETC pen-etration (�ETC Penetrationit), I now allow the effect of ETC tovary over time and across facilities as a function of the diffu-sion of ETC.19 As discussed, I must estimate equation (19) on

16. The scale of the graph is arbitrary. I set the omitted category to zero.Thus, for example, the log minimum toll in ETC Year 4 is 2∗β1. +2∗β3 and the logminimum toll in ETC Year −4 is 2∗β−4.

17. For the same reason, I do not extend the dummies in (18a) or (18b) formore years after ETC is adopted.

18. The point estimates in Figure IIIB indicate no preperiod trend in the bal-anced panel, which is reassuring relative to the (albeit statistically insignificant)suggestive evidence of some downward preperiod trend in the full sample in Fig-ure IIIA. In Table VI I investigate the issue of potential preperiod trends in moredetail, using a more parsimonious specification to increase statistical precision.

19. A more stringent test would be to include both �ETC Penetrationit andETCit on the right-hand side to examine whether the diffusion of ETC has animpact on toll rates that can be distinguished from a linear trend. I find thatwhile the two variables are jointly significant, it is not possible to distinguish theeffect of ETC penetration separately from a linear trend (not shown). This is notsurprising, because, on average, the data contain about six years of data on a



the subsample of facilities that never offer an ETC discount, aschanges in the ETC discount will affect both the diffusion of ETCand the minimum toll. Column (4) of Table IV shows the results.The coefficient on the change in the ETC penetration rate is 0.623(standard error 0.285). This indicates that every 10-percentage-point increase in ETC penetration is associated with a (statisti-cally significant) toll increase of 6.2%.

For the full sample of facilities, I estimate (19) instrumentingfor �ETC Penetrationit with the indicator variable ETCit; thisis equivalent to instrumenting for the change in ETC penetrationwith a linear trend. Column (5) shows these results. The coefficienton �ETC Penetrationit is 0.557 (standard error 0.262), indicatingthat every 10-percentage-point increase in ETC penetration isassociated with a (statistically significant) 5.6% increase in thetoll. To allow the effect of ETC to vary over time, in column (6) Iinstead instrument for the change in ETC penetration with a cubicpolynomial in the number of years the facility has had ETC. Thecoefficient on �ETC Penetrationit is now 0.501 (standard error0.261). The results are also similar if I instead instrument for�ETC Penetrationit with a series of indicator variables for thenumber of years under ETC (not shown).

The magnitude of the estimated effect of ETC is quite sim-ilar across all of the various specifications shown in Table IV.The results from the baseline specification (Table IV, column(1)) suggest that after 14 years, by which point ETC has dif-fused to its steady state level (see Figure II), ETC is associ-ated with an increase in the toll rate of 17%, or about one-sixth(∼exp(βETCAdopt + 14∗βETC)). The IV estimates in columns (5) and(6) suggest that once ETC has diffused to its steady state level of60%, it is associated with increases in tolls of 26 and 23%, respec-tively (∼exp(βETCAdopt + 0.6∗β�ETC Penetration)). When the sample islimited to facilities without ETC discounts, the implied steadystate increase in tolls is 36% when (3) is estimated (column (3))or 38% when (5) is estimated (column (4)). All of these impliedsteady state toll increases associated with ETC are statisticallysignificant at at least the 10% level. Taken together, these esti-mates suggest that the diffusion of ETC to its steady state levelis associated with a 20 to 40 percent increase in toll rates. Giventhe extremely inelastic demand for driving with respect to the toll

facility with ETC, and the diffusion pattern of ETC is basically linear for thosefirst six years (see Figure II).



that I estimate below, these results suggest that the associatedincrease in revenue for the toll authority is also about 20 to 40percent.

Endogeneity of the Timing of ETC Adoption. I have analyzedthe endogenous choice of tax rates while assuming that the choiceof the salience of the tax system (i.e., the adoption of ETC) isexogenous. In practice, the decision to adopt ETC does not appearto be random. For example, as previously discussed, higher laborcosts in the northeast may have encouraged more ETC adoption.This does not, however, pose a problem for the analysis per se,which requires only that the timing of ETC implementation beuncorrelated with changes in a facility’s toll setting relative to itsnorm.

Nonetheless, the correlation of various observable character-istics with whether or when a facility adopts ETC (see Table II)raises concerns about the identifying assumption that absent theintroduction of ETC on facility i in year t, toll rates would nothave changed differentially for that facility. I therefore analyzethe effect of ETC separately on samples stratified by these char-acteristics. Table V shows the results. Column (1) replicates thebaseline specification (Table IV, column (1)). Columns (2) through(7) show the effects separately by geographic region, by facilitytype (bridges and tunnels vs. roads), and by facility age. Not onlydoes statistical significance generally persist across the subsam-ples, but also the point estimates are remarkably similar.20 Tomore directly control for differences across facilities in the un-derlying rate of toll growth, column (8) shows that the resultsare robust to the addition of facility fixed effects to (17), which isequivalent to allowing facility-specific linear trends in toll rates.

One specific source of omitted variable bias that the precedinganalysis does not directly address is that ETC adoption may be apart of a broader infrastructure project, or a signal that infrastruc-ture modernization is in the works. In this case, the relationshipbetween ETC and toll increases may be spurious, as infrastruc-ture projects may necessitate (or provide political cover for) tollincreases. To investigate this possibility, I compiled histories of

20. As a distinct exercise, I was also interested in whether the impact of ETCvaried between operating authorities that automatically send monthly statementsof expenses to users and authorities from which drivers had to actively request(and in some cases pay for) ETC expense statements. The point estimates did notsuggest any economically or statistically differential impact of ETC on toll ratesalong this dimension, although the standard errors were sufficiently large so thatit was not possible to rule out fairly large differences (not shown).


E-ZTAX: TAX SALIENCE AND TAX RATES 999T

AB

LE

VIM

PA

CT

OF

ET

CO

NT

OL

LR

AT

ES:R

OB

US

TN

ES

SA

NA

LYS

IS

Nor

thea

stB

ridg

esO

pen

Ope

nF

acil

ity

Fac

ilit

ies

wit

han

dS

outh

and

afte

r19

60fi

xed

infr

astr

uct

ure

data

Bas

elin

em

idw

est

and

wes

tR

oads

tun

nel

s19

60or

befo

reef

fect

s(1

)(2

)(3

)(4

)(5

)(6

)(7

)(8

)(9

)(1

0)

ET

Cit

0.01

50.

016

0.01

40.

015

0.02

80.

021

0.01

30.

013

−0.0

55−0

.055

(0.0

06)

(0.0

10)

(0.0

05)

(0.0

08)

(0.0

10)

(0.0

10)

(0.0

07)

(0.0

07)

(0.0

35)

(0.0

35)

[.01

8][.

141]

[.03

0][.

067]

[.01

5][.

065]

[.07

9][.

072]

[.12

5][.

124]

ET

CA

dopt

it−0

.051

−0.0

48−0

.044

−0.0

23−0

.086

0.05

1−0

.084

−0.0

530.

014

0.01

4(0

.035

)(0

.054

)(0

.027

)(0

.063

)(0

.017

)(0

.079

)(0

.025

)(0

.036

)(0

.007

)(0

.007

)[.

158]

[.39

9][.

137]

[.71

9][.

000]

[.53

4][.

003]

[.14

7][.

048]

[.04

8]IN

FR

AA

DO

PT

it0.

017

(0.0

14)

[.22

1]IN

FR

Ait

−0.0

03(0

.007

)[.

659]

Mea

nde

p.va

r.0.

020

0.02

20.

017

0.02

10.

020

0.01

90.

021

0.02

00.

021

0.02

1#

ofst

ates

2414

1018

1613

2124

2323

#op

.au

thor

itie

s49

2821

2431

2039

4946

46#

offa

cili

ties

123

6855

4479

4377

123

115

115

N5,

079

3,00

82,

071

1,69

23,

387

1,38

93,

690

5,07

94,

712

4,71

2

Not

es.T

able

repo

rts

resu

lts

from

esti

mat

ing

vari

ants

of(1

7)by

OL

S.T

he

depe

nde

nt

vari

able

isth

ech

ange

inth

elo

gm

inim

um

toll

.All

regr

essi

ons

incl

ude

year

fixe

def

fect

s(n

otsh

own

).E

ach

oper

atin

gau

thor

ity

rece

ives

equ

alw

eigh

t.E

TC

Ado

ptit

isan

indi

cato

rva

riab

lefo

rw

het

her

faci

lity

iad

opte

dE

TC

inye

art.

ET

Cit

isan

indi

cato

rva

riab

lefo

rw

het

her

the

faci

lity

has

ET

C;i

tis

1in

the

year

that

ET

Cis

adop

ted

and

inal

lsu

bseq

uen

tye

ars.

Col

um

ns

(2)

and

(3)

lim

itth

esa

mpl

eto

,res

pect

ivel

y,fa

cili

ties

inth

en

orth

east

and

mid

wes

t,an

dfa

cili

ties

inth

eso

uth

and

wes

t.C

olu

mn

s(4

)an

d(5

)lim

itth

esa

mpl

eto

,res

pect

ivel

y,ro

ads,

and

brid

ges

ortu

nn

els.

Col

um

ns

(6)a

nd

(7)l

imit

the

sam

ple

to,r

espe

ctiv

ely,

faci

liti

esth

atop

ened

afte

r19

60an

dfa

cili

ties

that

open

edin

1960

orea

rlie

r.C

olu

mn

(8)

adds

faci

lity

fixe

def

fect

sto

the

righ

t-h

and

side

of(1

7).I

nco

lum

ns

(9)

and

(10)

the

sam

ple

isli

mit

edto

the

115

faci

liti

esfo

rw

hic

hin

fras

tru

ctu

reda

taar

eav

aila

ble.

INF

RA

Ado

ptit

isan

indi

cato

rva

riab

lefo

rw

het

her

faci

lity

ist

arte

da

new

infr

astr

uct

ure

proj

ect

inye

art.

INF

RA

itis

anin

dica

tor

vari

able

for

wh

eth

erfa

cili

tyi

has

anin

fras

tru

ctu

repr

ojec

tin

prog

ress

inye

art;

itis

1in

the

year

that

the

proj

ect

isst

arte

dan

din

alls

ubs

equ

ent

year

sth

atth

epr

ojec

tis

inpr

ogre

ss.A

lles

tim

ates

give

equ

alw

eigh

tto

each

oper

atin

gau

thor

ity.

Sta

nda

rder

rors

inpa

ren

thes

esar

ecl

ust

ered

byst

ate,

and

p-va

lues

are

show

nin

squ

are

brac

kets

.



infrastructure projects on 115 of the 123 individual toll facilities.21

These histories report the timing of a variety of infrastructureprojects including renovations, replacements, repairs, widenings,extensions, and other improvements. I constructed indicator vari-ables for whether facility i started an infrastructure project inyear t (INFRAAdoptit) and whether it had a project either startedor ongoing in year t (INFRAit). On average, a project was startedin 2.2% of facility-years, and 10.1% of facility-years had an infras-tructure project either starting or ongoing. I reestimate the basicrelationship between ETC and toll increases (equation (17)) withthese two additional variables included as covariates. Column (9)shows that the baseline results (without the additional infra-structure variables) are unaffected by restricting the sample tothe 115 facilities for which I have data on infrastructure projects.Column (10) shows that the estimated increase in tolls associatedwith ETC is not affected in either magnitude or statistical signif-icance by including the two infrastructure variables as controls.This suggests that the increase in tolls associated with ETC isnot likely to be spuriously due to a correlation between ETC andinfrastructure projects, which themselves are responsible for tollincreases; indeed, the results suggest that infrastructure projectsare not, in fact, associated with toll increases.

There are of course many reasons, besides infrastructureprojects, that the timing of ETC adoption might be spuriously cor-related with toll increases. For example, facilities may respond toincreased congestion by both adopting ETC and by raising tolls ascomplementary congestion-reducing strategies. This suggests weshould observe increases in congestion (or a proxy for it such astraffic) on a facility prior to ETC adoption. Alternatively, facili-ties might respond to a negative revenue shock by both raisingtolls and adopting ETC, with the latter a way to lower revenuelosses from the administrative costs of toll collection. This sug-gests we should observe declining revenue (or declining traffic) ona facility in the years prior to ETC adoption. More generally, wecan look for changes in toll rates in the years prior to ETC adop-tion as a partial test of the identifying assumption that absent theadoption of ETC, a facility would not have experienced differen-tial changes in its toll rate. Of course, if the lower salience of ETC

21. The primary source of data was facility Web pages and annual reports,which often provide detailed histories of work on the facilities. The level of detailand the nature of the projects reported vary across facilities. However, because allof the analysis is within-facility, this should not pose a problem.



TABLE VICHANGES IN TRAFFIC, REVENUE, AND TOLLS PRIOR TO ETC ADOPTION

Dep. var.: Dep. var.: Dep. var.:� log(traffic) � log(revenue) � log(minimum toll)

(1) (2) (3) (4) (5) (6)

1–2 years before −0.000 −0.009 0.004ETCAdoptedit (0.007) (0.016) (0.013)

[.955] [.599] [.777]1–5 years before 0.013 0.006 0.009

ETCAdoptedit (0.010) (0.012) (0.007)[.198] [.601] [.242]

ETCAdoptit −0.000 0.000 0.002 0.002 −0.051 −0.051(0.010) (0.010) (0.025) (0.025) (0.035) (0.035)

[.996] [.978] [.922] [.930] [.158] [.162]ETCit −0.006 −0.001 0.028 0.031 0.016 0.017

(0.010) (0.010) (0.015) (0.015) (0.006) (0.006)[.551] [.959] [.090] [.058] [.018] [.008]

Mean dep. var 0.049 0.077 0.020# of states 21 13 24# op. authorities 32 19 49# of facilities 76 45 123N 2,200 1,411 5,079

Notes. Table reports results from estimating variants of (17) by OLS. Dependent variables are definedin the column headings. In addition to the covariates shown in the table, all regressions include year fixedeffects. Each operating authority receives equal weight. Standard errors (in parentheses) are clustered bystate. p-values are reported in square brackets. “1–2 years before ETCAdoptedit” is an indicator variable forwhether it is one to two years before the facility adopts ETC. “1–5 years before ETCAdoptedit” is an indicatorvariable for whether it is one to five years before the facility adopts ETC. ETCAdoptit is an indicator variablefor whether facility i adopted ETC in year t. ETCit is an indicator variable for whether the facility has ETC;it is 1 in the year that ETC is adopted and in all subsequent years.

made it easier to raise tolls, ETC might be adopted precisely byfacilities that were encountering difficulties in making needed tollincreases, suggesting that facilities might experience declines intraffic, revenue, or toll increases prior to ETC adoption. Althoughevidence of such effects would therefore not necessarily be incon-sistent with the salience story, the lack of any such evidence re-duces concerns about omitted variable bias and spurious findings.

Table VI shows the results. I reestimate (17) with three dif-ferent dependent variables: �log(traffic)it (columns (1) and (2)),�log(revenue)it (columns (3) and (4)), and �log(minimum toll)it

(columns (5) and (6)). In addition to the standard regressors (yearfixed effects, ETCAdoptit, and ETCit), I also include an indicatorvariable for whether it is one to two years prior to ETC adop-tion (odd columns) or whether it is one to five years prior to ETC



adoption (even columns). The coefficients on these indicator vari-ables for years just prior to ETC adoption show no statistically orsubstantively significant evidence of systematic changes in traf-fic, revenue, or tolls in the years prior to a facility’s adoptingETC. These results are consistent with the results from estimat-ing (18a), which show no systematic preexisting trend in toll ratesprior to a facility’s adoption of ETC, particularly in the balancedpanel (see Figures IIIA and IIIB). One reason that the variousendogeneity concerns may not in practice be a problem is that, asnoted in Section IV.B, the different facilities run by a given oper-ating authority tend to adopt ETC all at the same time, and yetmay be experiencing different patterns of traffic and tolls.22

There are several other results of interest in Table VI. Thefinding in columns (3) and (4) that revenue increases by about 3percent per year under ETC is broadly consistent with the esti-mated increase in tolls under ETC and the finding that demandfor driving is very inelastic with respect to the toll.23 There is alsosome suggestive evidence in columns (1) and (2) that traffic de-clines under ETC, although these estimates are not statisticallysignificant and are substantively quite small; a decline in traf-fic would be consistent with the survey evidence in Section III ofoverestimation of toll levels by ETC users.

VI.B. The Impact of ETC on the Politics of Toll Setting

The model in Section II.B suggested two potential mecha-nisms behind a finding that reduced salience is associated withincreased tax rates: (i) a reduced behavioral responsiveness totaxes and (ii) a reduction in the political costs of tolls, particularlyin the differential political costs of tolls in election years com-pared to nonelection years. Section V presented evidence for thefirst potential mechanism. To investigate the political channel, Iexamine whether there are political costs to tolls and how thesecosts change under ETC.

Table VII shows the results. Because the political fallout fromraising tolls may be concentrated on the extensive margin (i.e.,whether tolls are raised), I report results not only for the baseline

22. In a different context, Dusek (2003) examines the impact of the introduc-tion of state income tax withholding on tax rates, but notes that the decision tointroduce income tax withholding appears to be correlated with increased demandfor bigger government, making the results hard to interpret.

23. For the sample for which I have revenue data, I estimate that ETC isassociated with a 2.2% increase in tolls each year (not shown).



TABLE VIITHE IMPACT OF ETC ON THE POLITICS OF TOLL SETTING

� log Min toll � log Min toll � log Min tollmin toll raised? min. toll raised? min. toll raised?

(1) (2) (3) (4) (5) (6)

ETCit 0.015 0.073 0.006 0.044 0.006 0.044(0.006) (0.024) (0.009) (0.022) (0.009) (0.022)

[.018] [.006] [.507] [.042] [.494] [.042]AnyElec −0.016 −0.029

Yearst (0.004) (0.010)[.000] [.003]

GovElec −0.016 −0.036Yearst (0.005) (0.012)

[.001] [.002]LegOnly −0.015 −0.021

ElecYearst (0.005) (0.012)[.005] [.085]

AnyElec 0.017 0.055Yearst (0.012) (0.027)*ETCit [.140] [.041]

GovElec 0.004 0.016Yearst (0.014) (0.033)*ETCit [.791] [.617]

LegOnly 0.030 0.094ElecYearst (0.014) (0.033)*ETCit [.038] [.005]

Notes. Columns (1) and (2) report estimates of (17); columns (3)–(6) report estimates of (20). Dependentvariable (shown in column heading) is �log minimum toll (odd columns) or an indicator variable for whetherthe minimum toll was raised (even columns). In addition to the covariates shown in the table, all regressionsinclude year fixed effects, ETCAdoptit , and interactions between ETCAdoptit and any indicator variables forthe election year included in the regression. Each operating authority receives equal weight. Standard errors(in parentheses) are clustered by state. p-values are in square brackets. “AnyElecYearst” is an indicatorvariable for whether state s’s governor or legislature is up for election in year t. “GovElecYearst” is anindicator variable for whether the governor (and therefore almost always the legislature as well) is up forelection. “LegOnlyElecYearst” is an indicator variable for whether only the legislature is up for election. ETCitis an indicator variable for whether the facility has ETC; it is 1 in the year that ETC is adopted and in allsubsequent years. Sample size in all columns is 5,079 facility-years, 123 facilities, 49 operating authorities,and 24 states. The mean of the dependent variable is 0.020 (odd columns) and 0.077 (even columns).

dependent variable � log minimum toll (odd columns) but alsofor the binary dependent variable of whether the minimum tollincreased (even columns). Column (1) replicates the baseline re-sults from (17) (see Table IV, column (1)). Column (2) shows theresults from estimating (17) with the binary dependent variablefor whether the minimum toll was raised that year; the coefficienton ETCit is 0.073 (standard error 0.024). This suggests that, rela-tive to the baseline 7.7% annual probability of a toll increase, theprobability of a toll increase almost doubles on a facility once it



has ETC. Combined with the evidence in column (1), this suggeststhat the increase in tolls associated with ETC comes about pri-marily through more frequent toll increases of similar magnitude.

I then expand the baseline specification in (17) to includeindicator variables for whether it is an election year, and the in-teractions of these indicators with the change in salience, as pro-posed in the estimating (15) from Section II.B. This allows me toexamine whether there is a political business cycle in toll settingand whether this political business cycle varies under manual tollcollection and ETC. Specifically, I estimate

yit = γt + β1ETCAdoptit + β2ETCit + β31(ElecYear)st

+β41(ElecYear)st ∗ ETCAdoptit

+β51(ElecYear)st ∗ ETCit + εit.(20)

Columns (3) and (4) report results when 1(ElecYear)st is an indi-cator for whether there is any state election (for either the gov-ernor or the legislature) in state s and year t; about half of thefacility-years in the data are election years, but the timing of theelectoral calendar varies across states. Columns (5) and (6) reportresults when 1(ElecYear)st is two separate indicators for whetherthe governor (and therefore almost always the legislature as well)is up for election and for whether only the legislature is up forelection; each of these indicator variables is turned on in roughlyone-fourth of state years.

In all four specifications, the coefficients on all of the elec-tion year indicators are negative and statistically significant; thisdemonstrates the political business cycle under manual toll col-lection. Given the average annual 2% increase in tolls, the coeffi-cient on the election year dummies of about −0.016 in columns (3)and (5) indicates that toll increases are about 75% lower duringelection years than during nonelection years under manual tollcollection.

The interaction term between the election year indicator vari-ables and ETC is always positive; it is statistically significant forlegislature-only election years (columns (5) and (6)) and statisti-cally significant (or only marginally insignificant) for any electionyear (columns (3) and (4)). This suggests that under ETC, toll-setting behavior is less sensitive to the political election calendar(particularly legislature elections) than under manual toll collec-tion. Indeed, there is no evidence that toll increases are lower inelection years relative to nonelection years under electronic toll



collection; the sum of the coefficients on the election year indicatorvariable and its interaction with ETC (i.e., β3 + β5) is almost al-ways positive (and never significantly negative).24

VII. ALTERNATIVE EXPLANATIONS

In this section, I briefly consider a range of alternative expla-nations for the increase in tolls associated with ETC other thanthe decline in the salience of the toll. I note at the outset thata general point in favor of the salience-based explanation is thefinding that toll setting becomes less sensitive to the local electioncalendar under ETC; this is consistent with a decline in saliencereducing the political costs of raising tolls, but would not be pre-dicted by any of the alternative explanations I discuss.

VII.A. ETC Lowers the Operating Cost of Toll Collection

ETC is associated with substantial reductions in the annualcosts of operating and maintaining toll facilities; the ETC cost sav-ings come primarily from reductions in the labor costs associatedwith manual toll collection (Hau 1992; Pietrzyk and Mierzejewski1993; Levinson 2002).25 However, for increases in the efficiency oftax collection to increase the equilibrium tax rate requires an im-provement in the marginal efficiency of tax collection (Becker andMulligan 2003). By contrast, ETC improves the fixed componentof the efficiency cost of taxation—because the administrative costsavings are independent of the toll rate—which should thereforenot prompt an increase in the rate of existing taxes.26

A decline in the fixed administrative costs of tax collectioncould, however, encourage the introduction of new taxes, such as

24. The “main effect” of ETC, although positive, is no longer statistically sig-nificant in columns (3) and (5); toll increases are not statistically significantlylarger in nonelection years under ETC than under manual toll collection. How-ever, toll increases are statistically significantly larger in election years underETC than under manual toll collection; the sum of the coefficients on ETC and theinteraction of ETC and election year (i.e., β2 + β5 in (20)) is statistically significantin column (3) and statistically significant for the legislative election year variablein column (5) (not shown).

25. Toll collection costs under manual toll collection can be quite high. A 1995study of turnpikes in Massachusetts and New Jersey estimated that toll collectioncosts under manual toll collection were about 6 percent of toll revenue (Friedmanand Waldfogel 1995); a 2006 study found that on portions of the MassachusettsTurnpike where there is relatively little traffic, toll collection costs were over one-third of toll revenue (Kriss 2006).

26. Note, moreover, that if operating authorities set tolls to meet an exogenousrevenue requirement, the reduction in administrative costs would lower (ratherthan raise) the equilibrium toll needed to raise a fixed amount of (net) revenue.



the introduction of tolls on roads that had not been previouslybeen tolled or the construction of new (tolled) roads where noroad existed before. Any such effects of ETC, however, would notshow up in my analysis, which limits the sample to facilities withpreexisting tolls. Lower fixed administrative costs of toll collectioncould also encourage the installation of more toll collection pointson an existing toll facility; however, I find no evidence that ETChad such an effect.27

VII.B. ETC Installation Requires Capital Outlay

Although ETC lowers the costs of operating and maintainingtoll facilities, installation of ETC requires a capital outlay. It seemsunlikely that this capital outlay would require an increase in tolls.Operating authorities can borrow to cover these capital costs, andthe capital costs are recouped within a few years by the savingsin operating and maintenance costs, and by revenue from the saleor lease of the transponders and interest on prepayments anddeposits (Hau 1992; Pietrzyk and Mierzejewski 1993). Of course,it is possible that operating authorities might use the installationcosts of ETC as an excuse to raise tolls, even though ETC is self-financing. Any such excuse might be used for a one-time increasein tolls when ETC comes in; it seems less natural that this excusecould be used for subsequent increases in tolls as ETC use diffusesamong drivers.

VII.C. Changes in Menu Costs Associated with ETC

It is possible that ETC lowers the administrative (menu) costof toll changes. There could be literal menu cost savings if signslisting the toll rate no longer had to be changed under ETC. Al-ternatively, ETC might allow smaller increases of non-“round”amounts; unlike manual tolls, this would not impose on driversthat they carry small coins. In practice, however, ETC tolls arenot less “round” than manual tolls, except when they are specifiedas a fixed percentage discount off of the manual toll. In addition,the increase in tolls associated with ETC persists for the sub-sample of facilities that do not offer discounts; for these facilities,there can be no menu cost savings, as changing the electronic toll

27. I reestimate (17) using as a dependent variable a binary measure forwhether there is an increase in the number of toll transactions someone driving aone-way, full-length trip on the facility would have to make. I perform this analysisfor the full sample of facilities, and separately both for roads and for bridges andtunnels (not shown).



requires changing the manual toll, and all facilities continue tohave at least some manual payers. Finally, even if ETC did reducemenu costs, this should suggest that ETC would be associatedwith more frequent toll adjustments, but it is not clear why thiswould produce a higher equilibrium toll rate.

VII.D. ETC Lowers Personal Compliance Costs of Toll Payment

ETC reduces the drivers’ compliance costs of paying tolls (Hau1992; Levinson 2002). Friedman and Waldfogel (1995) estimatethat under manual toll collection, these compliance costs—whichconsist of time spent queuing and paying tolls at the toll plaza—are, on average, about 15% of toll revenue. Reductions in compli-ance costs of paying tolls may directly increase drivers’ willing-ness to pay the (monetary) toll, and hence provide an alternativeexplanation for the observed increase in toll rates.

In practice, however, two independent pieces of empirical ev-idence suggest that toll authorities do not increase tolls in re-sponse to reductions in compliance costs; this is consistent withthe finding in Section V that they set tolls substantially below therevenue-maximizing rate (i.e., that they implicitly place a rela-tively large weight on consumer surplus). The first piece of sug-gestive evidence comes from variation across roads in the numberof times an individual must make a toll transaction, and hencevariation in the compliance costs savings from ETC. For example,in 1985 an individual made eleven toll transactions while drivingthe length of the Garden State Parkway, compared to only two onthe New Jersey Turnpike. If tolls were increased under ETC inresponse to the reductions in compliance costs, we would expectgreater toll increases on roads with a greater number of toll trans-actions. In fact, I find weak evidence of the opposite. The secondpiece of suggestive evidence comes from what happens to toll rateswhen a bridge or tunnel switches from collecting tolls at both endsof the facility to collecting tolls at only one end; at various timesover the course of my sample, about half of the bridges and tun-nels (40 of 79) made this switch, which reduced compliance costson their facility by one-half. I find little evidence of a substantivelyor statistically significant increase in tolls on a facility followingthis reduction in compliance costs.28

28. The results of both of these analyses are presented in more detail inthe Online Appendix (Section C) and in the working paper version of this paper(Finkelstein 2007).



VII.E. ETC Raises the Optimal Congestion-Correcting Toll

Could the increase in tolls under ETC come entirely from theincrease in the optimal congestion externality–reducing toll thatresults from the reduced consumer responsiveness to tolls? Thiswould suggest that the effect of ETC on toll rates is a salience ef-fect, but one that comes entirely from a reduction in salience at thetime of consumption (driving). This seems unlikely given the evi-dence in Section VI.B that ETC affects the political costs of raisingtolls; this suggests that at least some of the toll increase associ-ated with ETC is likely to be due to a decline in voting salience.In addition, as an (admittedly quite) crude test of whether theincrease in tolls under ETC is driven by an increase in congestionunder ETC, I experimented with controlling for traffic (a proxy forcongestion) on the right-hand side of (17). I found that the impactof ETC on the change in tolls is not sensitive to including traffic asa control, suggesting that even conditional on the level of traffic,tolls still rise under ETC (not shown).

VIII. CONCLUSIONS

This paper has examined the hypothesis that a less salienttax system can produce a higher equilibrium tax rate. Belief inthis possibility has contributed to opposition to tax reforms thatare believed to reduce tax salience, such as Federal income taxwithholding or partial replacement of the income tax with a value-added tax. Yet the sign of the effect of tax salience on tax rates istheoretically ambiguous, and empirical evidence has been lacking.

I examine the relationship between tax salience and tax ratesempirically by looking at the impact of electronic toll collection(ETC) on toll rates. Survey evidence indicates that drivers whopay tolls electronically are substantially less aware of toll ratesthan those who pay with cash, suggesting that ETC reduces tolls’salience. To analyze the impact of this reduction in salience, Iassembled a new data set on toll rates over the last half centuryon 123 toll facilities in the United States. Because different tollfacilities adopted ETC in different years, and some have not yetadopted it, I am able to examine the within-toll facility change intolls associated with the introduction of electronic toll collection.

I find robust evidence that toll rates increase following theadoption of electronic toll collection. The estimates suggest thatafter ETC use among drivers has diffused to its steady state level,



toll rates are 20 to 40 percent higher than they would have beenunder manual toll collection. I provide evidence of two poten-tial mechanisms by which reduced salience may contribute toincreased toll rates: under ETC driving behavior becomes lesselastic (in absolute value) with respect to the toll, and toll settingbecomes less sensitive to the local election calendar. This declinein the political costs of raising tolls associated with ETC wouldnot be predicted by alternative explanations for the increase intolls associated with ETC. I also present additional evidence thatis not consistent with specific alternative explanations.

As previously discussed, the normative implications of thesefindings are ambiguous. Evidence on what is done with the ex-tra revenue from the higher tolls—in particular, whether it isused for purposes that may be valued by users of the facilitysuch as infrastructure investment or reductions in other high-way fees, or whether it primarily serves to increase rents for thegoverning authority through increased employment or salaries ofbureaucrats—could help shed some light on the normative impli-cations of the higher tolls under ETC. Unfortunately, the availabledata are not sufficient for analysis of this issue.

The results also leave open the question of how tax salienceaffects tax rates in other contexts, such as federal income tax with-holding or the replacement of a sales tax with a value added tax.As previously discussed, the sign of the effect of tax salience on taxrates may well differ for taxes that are a larger share of expendi-tures than tolls. The magnitude of any effect of tax salience is alsolikely to differ across different political institutions. The resultsin this paper suggest that the salience of the tax instrument is animportant element to consider in both theoretical and empiricalinvestigations of the political economy of tax setting. Relatedly,they suggest that the empirical impact of tax salience in theseother specific settings is an interesting and important directionfor further work.

MASSACHUSETTS INSTITUTE OF TECHNOLOGY AND NBER

REFERENCES

Becker, Gary, and Casey Mulligan, “Deadweight Costs and the Size of Govern-ment,” Journal of Law and Economics, 46 (2003), 293–340.

Brennan, Geoffrey, and James Buchanan, The Power to Tax: Analytical Founda-tions of a Fiscal Constitution (Cambridge, UK: Cambridge University Press,1980).



Buchanan, James, Public Finance in Democratic Process: Fiscal Institutions andthe Individual Choice (Chapel Hill: University of North Carolina Press, 1967).

Buchanan, James, and Richard E. Wagner, Democracy in Deficit (New York:Academic Press, 1977).

Chetty, Raj, Kory Kroft, and Adam Looney, “Salience and Taxation: Theory andEvidence,” American Economic Review, forthcoming.

Dollery, Brian, and Andrew Worthington, “The Empirical Analysis of Fiscal Illu-sion,” Journal of Economic Surveys, 10 (1996), 261–298.

Dusek, Libor, Do Governments Grow When They Become More Efficient? Evi-dence from Tax Withholding, Unpublished Ph.D. Dissertation, University ofChicago, 2003.

Feldman, Naomi, and Peter Katuscak, “Should the Average Tax Rate Be Marginal-ized?” mimeo, Ben Gurion University, 2005.

Finkelstein, Amy, “E-Z Tax: Tax Salience and Tax Rates,” Working Paper 12924,National Bureau of Economic Research, 2007.

Friedman, David, and Joel Waldfogel, “The Administrative and Compliance Costof Manual Highway Toll Collection: Evidence from Massachusetts and NewJersey,” National Tax Journal, June 1995.

Friedman, Milton, and Rose Friedman, Two Lucky People (Chicago: University ofChicago Press, 1998).

Hau, Timothy, “Congestion Charging Mechanisms for Roads: An Evaluation ofCurrent Practice,” Research Paper 1071, The World Bank, 1992.

Holguin-Veras, Jose, Ozbay Kaan, and Allison de Cerrano, “Evaluation Study of thePort Authority of New York and New Jersey’s Time of Day Pricing Initiative,”report, Rensselaer Polytechnic Institute, 2005.

Kearney, Melissa, “The Economic Winners and Losers of Legalized Gambling,”National Tax Journal, 58 (2005), 281–302.

Kriss, Eric, “Turnpike Task Force Final Report,” Pioneer Institute board presen-tation by Eric Kriss, October 18, 2006.

Levinson, David, Financing Transportation Networks (Northampton, MA: EdwardElgar, 2002).

Liebman, Jeffrey, and Richard Zeckhauser, “Schmeduling,” unpublished mimeo,Harvard’s Kennedy School of Government, 2004).

Nordhaus, William, “The Political Business Cycle,” Review of Economic Studies,42 (1975), 169–190.

Oates, Wallace, “On the Nature and Measurement of Fiscal Illusion: A Survey,”in Taxation and Fiscal Federalism: Essays in Honour of Russell Mathews,Geoffrey Brennan, Bhajan Grewal, and Peter Groenewegen, eds. (Sydney:Australian National University Press, 1988).

Peltzman, Sam, “Pricing in Public and Private Enterprises: Electric Utilities inthe United States,” Journal of Law and Economics, 14 (1971), 109–147.

Pietrzyk, Michael, and Edward Mierzejewski, “Electronic Toll and Traffic Man-agement Systems” (Washington, DC: National Academy Press, 1993).

Slemrod, Joel, “Which Is the Simplest Tax System of Them All?” in The Economicsof Fundamental Tax Reform, Hank Aaron and William Gale, eds. (Washington,DC: The Brookings Institution, 1996).

Soman, Dilip, “Effects of Payment Mechanism on Spending Behavior: The Role ofRehearsal and Immediacy of Payments,” Journal of Consumer Research, 27(2001), 460–474.

Thaler, Richard H., “Mental Accounting Matters,” Journal of Behavioral DecisionMaking, 12 (1999), 183–206.

The President’s Advisory Panel on Federal Tax Reform, Simple, Fair, and Pro-Growth: Proposals to Fix America’s Tax System (Washington, DC: U.S. Gov-ernment Printing Office, 2005).

U.S. Census Bureau, “Annual Survey of State and Local Government Finances andCensus of Governments,” 1985.

U.S. Department of Transportation, “Highway Statistics, YEAR,” various years.Wooldridge, Jeffrey, Econometric Analysis of Cross Section and Panel Data (Cam-

bridge, MA: MIT Press, 2002).


: tax salience and tax rates

Documents